The present paper shows that the term “complexity” includes absolutely different notions than now it seems to be presented in modern science and philosophy. V.S. Stepin’s postnon-classics has come to this new recognition too close, but, actually, it is a new recognition of uncertainty for systems of the third type (not deterministic and not stochastic). We introduce the interpretation of a type I uncertainty that implies that stochastic methods show systems identified, but methods of the theory of chaos and self-organization and neurocomputing show significant difference of target systems (processes). The concrete examples show the type I uncertainty and give an idea of a type II uncertainty, that implies the coincidence of distribution functions f(x) for different samplings. We prove that neurocomputing method not only differentiates samplings, but also identifies order parameters. In this case we also solve the system synthesis problem.